agentlenspro-visualize-context
Measure what a Claude Code agent actually puts in its context window, and what its SECOND turn costs — by spawning it for real and reading the raw API captures, not by guessing from its definition. Load this when someone asks how a new or custom agent compares to the built-in Explore / Plan / general-purpose agents, whether an agent is "lean", why a subagent is expensive, what CLAUDE.md or the MCP schemas or the skill listing cost inside an agent, or whether an agent's prompt cache survives from one turn to the next. Produces a measured element-by-element breakdown of turn 1 and turn 2, the exact diff between them, and a cache verdict that is predicted from the request bytes and then checked against what the API actually billed. Every number comes from count_tokens; nothing is estimated. Invoked as /agentlenspro-visualize-context <agent-name> [<agent-name> ...].
Source facts
- Repository
- Emasoft/AgentlensPro
- Last source activity
- July 30, 2026 at 15:42
- Detected SKILL.md language
- English
- Stars
- 3
- Forks
- 0
Install options
The review-first prompt is selected by default. You can switch to a direct command or download a local copy.
Review the source files
Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.
Showing SKILL.md
| name | agentlenspro-visualize-context |
|---|---|
| description | Measure what a Claude Code agent actually puts in its context window, and what its SECOND turn costs — by spawning it for real and reading the raw API captures, not by guessing from its definition. Load this when someone asks how a new or custom agent compares to the built-in Explore / Plan / general-purpose agents, whether an agent is "lean", why a subagent is expensive, what CLAUDE.md or the MCP schemas or the skill listing cost inside an agent, or whether an agent's prompt cache survives from one turn to the next. Produces a measured element-by-element breakdown of turn 1 and turn 2, the exact diff between them, and a cache verdict that is predicted from the request bytes and then checked against what the API actually billed. Every number comes from count_tokens; nothing is estimated. Invoked as /agentlenspro-visualize-context <agent-name> [<agent-name> ...]. |